2 months ago
Responsibilities
- Develop and improve object-detection and multi-object-tracking models for vehicles, pedestrians, and other road users.
- Own data strategy, including sampling, labeling, dataset versioning, hard-negative mining, class-imbalance handling, edge-case analysis, and deployed-system feedback.
- Design rigorous experiments and ablations while identifying overfitting, leakage, noisy labels, and improvements that do not survive field deployment.
- Build reproducible training pipelines with experiment tracking, configuration management, artifact lineage, model registries, metric dashboards, and automated hyperparameter search.
- Define product-relevant evaluation covering precision/recall trade-offs, class and scenario slices, calibration, and tracking quality.
- Validate models on deployed or field-collected data and manage the cycle from failure discovery through retraining and verified improvement.
- Partner with embedded engineers to optimize models for edge deployment within accuracy, latency, memory, and power constraints.
- Set technical direction, lead design reviews, mentor engineers, and improve standards for ML rigor, reproducibility, and production readiness.
Requirements
- 7+ years of experience in machine learning or computer vision with a record of shipping and improving production models.
- Deep hands-on experience with object detection and multi-object tracking, including modern architectures, data augmentation, and evaluation methods.
- Strong understanding of class imbalance, overfitting, dataset leakage, label noise, domain shift, model calibration, and statistically sound validation.
- Experience building training infrastructure or platforms for reproducible experiments, distributed training, hyperparameter search, metric comparison, and model lineage.
- Strong Python and PyTorch skills and practical experience with large image and video datasets.
- Experience validating models on deployed or field-collected data and owning model improvement from failure discovery through retraining.
- Demonstrated technical leadership across ambiguous, cross-functional work with clear communication and strong ownership.
- Preferred experience includes traffic, automotive, robotics, surveillance, video analytics, re-identification, trajectory modeling, occlusion handling, camera calibration, multi-camera tracking, active learning, weak supervision, synthetic data, automated labeling, or dataset-quality tooling.
- Preferred experience optimizing and deploying vision models on NVIDIA Jetson, Qualcomm Snapdragon, or other edge accelerators.
Benefits
- Work on computer vision intended to improve road safety and help save lives.
- Small, experienced team where senior engineers have meaningful technical influence.
- Fast-moving Series A startup environment with meaningful ownership and early-stage equity.
- Competitive compensation.
Categories
About Obvio
Obvio builds solar-powered, AI-assisted camera systems and software that automate traffic-law enforcement at pedestrian-heavy locations for cities and transportation agencies. The company generates revenue by deploying and operating its hardware-plus-software platform under contracts with municipalities. Headquartered in San Carlos, California, this privately held firm reports that early partner cities saw a 50% reduction in reckless driving after deployment.